http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47274
--- Comment #8 from Jan Hubicka <hubicka at ucw dot cz> 2011-01-17 16:03:37 UTC --- > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=47274 > > --- Comment #7 from John David Anglin <danglin at gcc dot gnu.org> 2011-01-17 > 15:42:26 UTC --- > Resolution file is: > > 3 > abs-1.o 3 > 70 262910e5 PREEMPTED_IR main_test > 84 262910e5 PREEMPTED_IR abs > 88 262910e5 PREEMPTED_IR abs_called > abs-1-lib.o 3 > 70 b13b015b PREVAILING_DEF_IRONLY abs > 101 b13b015b PREEMPTED_IR inside_main > 103 b13b015b PREVAILING_DEF_IRONLY abs_called > main.o 3 > 70 e5772d37 PREVAILING_DEF main > 76 e5772d37 PREVAILING_DEF_IRONLY main_test > 79 e5772d37 PREVAILING_DEF_IRONLY inside_main > > /tmp/ccJGcG05.lto.o is empty: > dave@gsyprf11:~/gcc-4.6/objdir/gcc/testsuite/gcc$ ls -l /tmp/ccJGcG05.lto.o > -rw------- 1 dave users 0 Jan 17 07:20 /tmp/ccJGcG05.lto.o > > Assert fails at this point: > > Breakpoint 1, lto_varpool_replace_node (prevailing_node=0x404591c0, > vnode=0x40459150) at ../../gcc/gcc/lto-symtab.c:306 > 306 gcc_assert (!vnode->analyzed || prevailing_node->analyzed); > (gdb) p vnode->analyzed > $7 = 1 > (gdb) p prevailing_node->analyzed > $8 = 0 We need to debug how the defined node ends up to be unanalyzed. I assume it is abs_called variable? It seems that we get wrong already when streaming abs-1-lib.o file. Would be possible to attach cgraph dump from WPA? Honza
